ARMA 2: OA beta build 86055

[CarlGustaffa 4]

Played a 12 hour long session today. Not many players, but overall I got pretty good performance. Very smooth. Some weird and bizarre AI sliding at distance, but much better than warping ones. Other than that, not a glitch, bug, crash, or the remotest of hickups. No malloc assignment used. Other than some chopper ferrying, played only as infantry. Haven't tried the campaigns yet though.

[themaster303 22]

does -malloc=tbb4malloc_bi also effect the performence of the server ??

[Dwarden 1124]

does -malloc=tbb4malloc_bi also effect the performence of the server ??

http://forums.bistudio.com/showpost.php?p=2048238&postcount=47

[jt445 10]

http://forums.bistudio.com/showpost.php?p=2048238&postcount=47

Can we get better multi-core dedicated server performance in 1.61? AI processing does not seem to be multi-threaded.

[stk2008 14]

Can we get better multi-core dedicated server performance in 1.61? AI processing does not seem to be multi-threaded.

This would be great to see dedi servers making use of multiple cores :)

[RSF TheCapulet 59]

As a developer, this is even my point of view. I gave up trying to understand inner workings of memory allocators, because it is too much complex area for me. I prefer "evidence based" approach - the allocator is almost a black box for me, and the one giving best results in empirical testing is the one I will prefer. The allocator crashing often is one I will not use. If you want to understand the differences, be prepared to face some difficult science.

Related: I also no longer have an ambition to create my own allocator from the scratch. This might have sense in OFP/Arma times, but those days are gone.

Have you experimented with more than just one type of allocators, like ptmalloc? (as others like malloc and mtmalloc both have terrible core scaling) And not knowing if it's feasible for game engine design, have you considered memory pools over malloc calls?

Edited November 5, 2011 by TheCapulet

[Dwarden 1124]

Have you experimented with more than just one type of allocators, like ptmalloc? (as others like malloc and mtmalloc both have terrible core scaling) And not knowing if it's feasible for game engine design, considered memory pools over an allocator?

keep in mind what Suma posted here:

http://forums.bistudio.com/showpost.php?p=2044253&postcount=92

We have used several memory allocators written by experts, including TCMalloc, NedMalloc and jemalloc. Most of them broke down under the ArmA 2 load, causing memory corruption or crashes.